Add PTY-byte capture tooling for RefTest scenarios #114

Open
opened 2026-09-05 00:20:06 +09:00 by natsukium · 0 comments
Owner

Migrated from docs/backlog.md (Open / RefTest capture tooling).

The replay harness exists (crates/felis-grid/tests/ref_recordings.rs: recording.bin + size.json per tests/ref/<scenario>/ directory, screen pinned as an insta snapshot), but the two seed scenarios are hand-authored. A capture tool would let any two-minute demonstration become a fixture. A felis-client --record flag is the wrong tap for this architecture: the client never sees raw PTY bytes (the daemon parses), so the recorder must tap the daemon's PTY read loop (likely a debug flag or env var writing the verbatim byte stream per session) or wrap the capture in a standalone PTY tee. Decide the surface, then grow scenarios from the deferred esctest clusters.

Design choice: daemon PTY-read tap vs standalone PTY tee. Once decided, add the recorder and seed additional tests/ref/<scenario>/ fixtures.

Origin: docs/backlog.md RefTest capture tooling.

Migrated from `docs/backlog.md` (Open / RefTest capture tooling). > The replay harness exists (`crates/felis-grid/tests/ref_recordings.rs`: `recording.bin` + `size.json` per `tests/ref/<scenario>/` directory, screen pinned as an insta snapshot), but the two seed scenarios are hand-authored. A capture tool would let any two-minute demonstration become a fixture. A `felis-client --record` flag is the wrong tap for this architecture: the client never sees raw PTY bytes (the daemon parses), so the recorder must tap the daemon's PTY read loop (likely a debug flag or env var writing the verbatim byte stream per session) or wrap the capture in a standalone PTY tee. Decide the surface, then grow scenarios from the deferred esctest clusters. Design choice: daemon PTY-read tap vs standalone PTY tee. Once decided, add the recorder and seed additional `tests/ref/<scenario>/` fixtures. Origin: `docs/backlog.md` RefTest capture tooling.
Sign in to join this conversation.
No description provided.